[Feature Request] RandomUniform operator support float16

Describe the feature request

https://onnx.ai/onnx/operators/onnx__RandomUniformLike.html#l-onnx-doc-randomuniformlike
☝️ as document said, Onnx RandomUniform op should support float16 dtype.

Describe scenario use case

Question

I attempted to generate a graph utilizing a RandomUniform node with an input data type of float16, and attempted to conduct inference using the ONNX API. However, when invoking the onnxruntime.InferenceSession, I encountered a NotImplemented error. The relevant code snippet is provided below.

Further information

• Version of onnx: 1.15.0

• Version of onnxruntime: 1.16.1

• Relevant Area: model usage, operators, IR

• Is this issue related to a specific model?
  No

Notes

import onnx
from onnx import helper
from onnx import TensorProto
import onnxruntime 

# The protobuf definition can be found here:
# https://github.com/onnx/onnx/blob/master/onnx/onnx.proto

dtype = TensorProto.FLOAT16
# Create one output (ValueInfoProto)
Y1 = helper.make_tensor_value_info('Y1', dtype, [3, 4])

# Create a node (NodeProto) - This is based on Pad-11
random_uniform_node = helper.make_node(
    'RandomUniform', # node name
    [], # inputs
    ['Y1'], # outputs
    name="random_uniform_node",
    dtype=dtype,
    high=1.0,
    low=0.0,
    seed=0.5,
    shape=[3, 4] # attributes
)
print(random_uniform_node)

# Create the graph (GraphProto)
graph_def = helper.make_graph(
    [random_uniform_node],
    'test-model',
    [],
    [Y1],
)

# Create the model (ModelProto)
model_def = helper.make_model(graph_def, producer_name='onnx-example')

print('The model is:\n{}'.format(model_def))
onnx.checker.check_model(model_def)
print('The model is checked!')

onnx.save(model_def, 'random_uniform.onnx') 

sess = onnxruntime.InferenceSession('random_uniform.onnx') 

output = sess.run(['Y1'], {})[0] 
print(Y1)
print(output)

error shown as:

NotImplemented: [ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation for RandomUniform(1) node with name 'random_uniform_node'

Start with the RandomUniform operator support used by onnxruntime.InferenceSession and reproduce the failure with the supplied Python example using TensorProto.FLOAT16. Confirm that the model loads and inference succeeds for a RandomUniform node with float16 output, without the reported NotImplemented error.